Yes, banks hold some quantity of treasuries at almost any time (and recent Basel rules oblige them to do so). It's important to note that treasuries carry price risk, i.e. they can decline in price and bank would suffer, and the longer the maturity, the higher the price risk. The reserve position at Fed doesn't have this risk, it's just cash. That's why banks may prefer reserves in certain situations.
The Fed doesn't need assets to create money, it acquires (buys) assets when it creates new money. It's needed to be noted that when the Fed creates money it doesn't add any financial assets (wealth) to the economy, only changes the composition. When the Fed posts operating profit, most of it goes to Treasury. When the Fed posts loss, it's usually also offset by capital injection by Treasury.
Treasury securities are the perfect collateral for the markets. When the Fed buys bonds under QE, it does so for market prices. But it's also possible (and was possible long before the crisis) to sell them to the Fed at a small discount and get reserves in return (so called discount window). This combined with Fed's repo operations makes reserves virtually unlimited and reserve requirements obsolete.
You also have to distinguish between reserves and capital. Reserves can only be used in transactions between banks and Treasury, be converted to cash and be exchanged for treasuries. These operations can't help if a bank has capital shortfall. Capital can only be replenished with external sources (like bailout by Treasury).
Another crisis event will likely cause deflation because of lending freeze, unemployment rise and incomes fall. These consequences together will cause demand slump and in turn price deflation. The alternative option is stagflation, but it's hard to predict.
Thanks. I'm going to try regurgitating my understanding with some new assumptions and hope I'm getting close to truth.
-So banks don't get to sell T-bonds at face value, just market price or sometimes even below for cash. The Fed allows this both to prevent problems with reserve requirements (and everything going along with banks not having enough) and I'd guess to also prevent large dumps on the T-bond market, which in turn makes T-bonds relatively stable, thus more attractive.
Many large banks are members of the Fed (paying both dues and fees), but the profit goes to the Treasury. The Treasury does create money, but it doesn't create wealth -- banks don't get
free wealth from the Fed, while benefits are all fairly well-balanced to keep money-creation powers not fully divested to either the market or the government, permitting a semi-market-based approach to money creation. Since Fed profits go to the Treasury (less overhead, losses, etc), the Treasury is effectively allowed to issue T-bonds and not pay interest, but only for T-bonds sold to the Fed at or below market rates. Discounting debt devaluation (in real terms) from inflation, this scheme of doing things prevents the government from just printing any financial troubles away (if they tried, the uncertainty from breaking this unspoken contract risks complete USD and T-bond meltdown, which could bring a ruinous banking crisis). In times of banking turbulence, like now and in recent history, lots of money is created this way, but banks use this money generally to meet reserve requirements (either those set by gov't or those self-imposed) after taking losses or otherwise over-extending themselves, so it doesn't just go right back into the economy, at least not until the banks recover and a healthy (or dangerously over-permissive, maybe "predatory") state of lending resumes. Everything works in a very complex, somewhat balanced way that usually (except in cases of extreme political pressure?) prevents quick and extreme reactions to short-term or mid-term problems. This permits long-term health for the US dollar.
I'm still not sure why M2 increase doesn't eventually mean we'll feel inflation from it when lending fully un-thaws, especially if the effects are suppressed/delayed by low money velocity. It doesn't factor in debts written off by banks for things like mortgages or LoCs, so maybe that acts as a deflationary force, reducing money velocity (lower home prices, less money circulating, especially lent money) and reducing the money supply (houses aren't counted in M2, but banks do have to write off the debts occasionally and usually take large losses in trying to sell the mortgaged property - consumers and businesses taking LoC probably aren't buying CDs, T-bonds, or making demand deposits). -But the housing market's had a huge uptick over the last couple years, unemployment's stabilized, and real wages are on the way up, so if M2 had any impact at all, why hasn't it effectively turned the value of our dollars to ash? Conservative banks?
(side-question: is money for repos also created or even significant compared to Fed T-bond purchases?)